
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import sklearn
import itertools
from sklearn.preprocessing import MinMaxScaler
from sklearn.neighbors import KNeighborsRegressor
from sklearn.linear_model import LinearRegression, LogisticRegressionCV
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import r2_score, mean_absolute_error, mean_squared_error, accuracy_score, roc_curve
from sklearn.model_selection import train_test_split, cross_val_score
from pandas import DataFrame
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import Input
from tensorflow.keras import layers
from sklearn.model_selection import GridSearchCV


df_bp=pd.read_excel('X_bp.xlsx', index_col=0)
df_bp


df_nup=pd.read_excel('X_nup.xlsx', index_col=0)
df_nup


df = df_bp.merge(df_nup, left_index=True,right_index=True, how='inner')

df


df.describe()


df.duplicated().sum()

0


df.info()

<class 'pandas.core.frame.DataFrame'>
Float64Index: 1023 entries, 0.0 to 1022.0
Data columns (total 13 columns):
 #   Column                                Non-Null Count  Dtype  
---  ------                                --------------  -----  
 0   Соотношение матрица-наполнитель       1023 non-null   float64
 1   Плотность, кг/м3                      1023 non-null   float64
 2   модуль упругости, ГПа                 1023 non-null   float64
 3   Количество отвердителя, м.%           1023 non-null   float64
 4   Содержание эпоксидных групп,%_2       1023 non-null   float64
 5   Температура вспышки, С_2              1023 non-null   float64
 6   Поверхностная плотность, г/м2         1023 non-null   float64
 7   Модуль упругости при растяжении, ГПа  1023 non-null   float64
 8   Прочность при растяжении, МПа         1023 non-null   float64
 9   Потребление смолы, г/м2               1023 non-null   float64
 10  Угол нашивки, град                    1023 non-null   float64
 11  Шаг нашивки                           1023 non-null   float64
 12  Плотность нашивки                     1023 non-null   float64
dtypes: float64(13)
memory usage: 111.9 KB


for col in df.columns: 
  plt.figure(figsize=(15, 5))
  plt.title("Гистограмма "+str(col))
  sns.histplot(data=df[col])
  plt.show()


for col in df.columns: 
    sns.catplot(y=col, data=df, kind='box')
    plt.title(col, fontsize=15)


cols = df.columns
g = sns.PairGrid(df[cols])
g.map(sns.scatterplot)

<seaborn.axisgrid.PairGrid at 0x2c0ae885ee0>


corr = df.corr()
mask = np.triu(np.ones_like(corr, dtype=bool))

# Создаем полотно для отображения большого графика
f, ax = plt.subplots(figsize=(15, 15))

# Создаем цветовую палитру
cmap = sns.diverging_palette(15, 3, as_cmap=True)

# Визуализируем данные кореляции 
sns.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})

<AxesSubplot:>


#Составим список признаков, у которых более 95% строк содержат одно и то же значение.
num_rows = len(df.index)
low_information_cols = [] #

for col in df.columns:
    cnts = df[col].value_counts(dropna=False)
    top_pct = (cnts/num_rows).iloc[0]
    
    if top_pct > 0.95:
        low_information_cols.append(col)
        print('{0}: {1:.5f}%'.format(col, top_pct*100))
        print(cnts)
        print()


# очистим данные от выбросов
for x in df.columns:
    q75,q25 = np.percentile(df.loc[:,x],[75,25])
    intr_qr = q75-q25

    max = q75+(1.5*intr_qr)
    min = q25-(1.5*intr_qr)

    df.loc[df[x] < min,x] = np.nan
    df.loc[df[x] > max,x] = np.nan


# проверим какое количество выбросов по каждому столбцу
df.isnull().sum()

Соотношение матрица-наполнитель          6
Плотность, кг/м3                         9
модуль упругости, ГПа                    2
Количество отвердителя, м.%             14
Содержание эпоксидных групп,%_2          2
Температура вспышки, С_2                 8
Поверхностная плотность, г/м2            2
Модуль упругости при растяжении, ГПа     6
Прочность при растяжении, МПа           11
Потребление смолы, г/м2                  8
Угол нашивки, град                       0
Шаг нашивки                              4
Плотность нашивки                       21
dtype: int64


# количество выбросов довольно мало, можно просто удалить эти строки
df = df.dropna(axis = 0)


# повторно анализируем диаграммы размаха
for col in df_bp.columns: 
    sns.catplot(y=col, data=df_bp, kind='box')
    plt.title(col, fontsize=15)


scaler = MinMaxScaler()
norm_df=pd.DataFrame(scaler.fit_transform(df),
            columns=df.columns, index=df.index) 
norm_df.describe()


norm_df.describe()

df


target = norm_df['Прочность при растяжении, МПа']
train = norm_df[[ 'модуль упругости, ГПа']]


cols=['Прочность при растяжении, МПа','модуль упругости, ГПа']
g = sns.PairGrid(df[cols])
g.map(sns.scatterplot)

<seaborn.axisgrid.PairGrid at 0x2c0b6f216a0>


X = df.copy()
X


from sklearn.model_selection import train_test_split
Xtrain, Xtest, Ytrain, Ytest = train_test_split(train, target, test_size=0.3, random_state=1)
Xtrain


from sklearn.linear_model import LinearRegression, LogisticRegression
lr = LinearRegression()
lin_reg_mod = LinearRegression()
lin_reg_mod.fit(Xtrain, Ytrain)
pred1 = lin_reg_mod.predict(Xtrain)
pred2 = lin_reg_mod.predict(Xtest)

from sklearn.metrics import mean_absolute_error, mean_squared_error, median_absolute_error, r2_score
print('MSE train: {:.3f}, test: {:.3f}'.format(
        mean_squared_error(Ytrain, pred1),
        mean_squared_error(Ytest, pred2)))
print('R^2 train: {:.3f}, test: {:.3f}'.format(
        r2_score(Ytrain, pred1),
        r2_score(Ytest, pred2)))

MSE train: 0.036, test: 0.035
R^2 train: 0.001, test: -0.008


all_accuracies = cross_val_score(estimator=lr, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.019900774405664623


knn = KNeighborsRegressor(n_neighbors=5)
knn.fit(Xtrain, Ytrain)
y_pred_knn = knn.predict(Xtest)
print("Оценка R2 к-ближайших соседей:", r2_score(Ytest, y_pred_knn))

Оценка R2 к-ближайших соседей: -0.2809276120596824


all_accuracies = cross_val_score(estimator=knn, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.2397751063721448


from sklearn.ensemble import RandomForestRegressor
rfr = RandomForestRegressor()
model = RandomForestRegressor(n_estimators=1500, max_depth=10)
model.fit(Xtrain, Ytrain)
y_pred_forest = model.predict(Xtest)
max_features: ['auto','sqrt','log2']
print('R^2 test: {:.3f}'. format(r2_score(Ytest, y_pred_forest)))

R^2 test: -0.166


all_accuracies = cross_val_score(estimator=rfr, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.5394410777074888


Xtrain, Xtest, Ytrain, Ytest = train_test_split(train, target, test_size=0.3, random_state=1)
Xtrain


print("Размер тренировочного датасета на входе:", Xtrain.shape)
print("Размер тестового датасета на входе:", Xtest.shape)
print("Размер тренировочного датасета на выходе:", Ytrain.shape)
print("Размертестового датасета на выходе:", Ytest.shape)

Размер тренировочного датасета на входе: (655, 1)
Размер тестового датасета на входе: (281, 1)
Размер тренировочного датасета на выходе: (655,)
Размертестового датасета на выходе: (281,)


lr = LinearRegression()
lr.fit(Xtrain, Ytrain)
y_pred = lr.predict(Xtest)
print("Оценка R2 линейная регрессия:", r2_score(Ytest, y_pred))

Оценка R2 линейная регрессия: -0.008008096662396769


all_accuracies = cross_val_score(estimator=lr, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.019900774405664623


knn = KNeighborsRegressor(n_neighbors=5)
knn.fit(Xtrain, Ytrain)
y_pred_knn = knn.predict(Xtest)
print("Оценка R2 к-ближайших соседей:", r2_score(Ytest, y_pred_knn))

Оценка R2 к-ближайших соседей: -0.2809276120596824


all_accuracies = cross_val_score(estimator=knn, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.2397751063721448


model = RandomForestRegressor(n_estimators=2000, max_depth=5)
model.fit(Xtrain, Ytrain)
y_pred_forest = model.predict(Xtest)
print('R^2 test: {:.3f}'. format(r2_score(Ytest, y_pred_forest)))

R^2 test: -0.028


all_accuracies = cross_val_score(estimator=rfr, X=Xtrain, y=Ytrain, cv=10)
print(all_accuracies.mean())

-0.548077758704397


from sklearn.model_selection import GridSearchCV


random_forest_tuning = RandomForestRegressor(random_state = 30)
param_grid = {
   'n_estimators': [100, 200, 500,700,900],
   'max_features': ['auto', 'sqrt', 'log2'],
   'max_depth' : [2,3,],
   'criterion' :['squared_error']
}
GSCV = GridSearchCV(estimator=random_forest_tuning, param_grid=param_grid, cv=5, verbose=2)
GSCV.fit(Xtrain, Ytrain)
GSCV.best_params_

Fitting 5 folds for each of 30 candidates, totalling 150 fits
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=2, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=auto, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=sqrt, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=100; total time=   0.0s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=200; total time=   0.1s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=500; total time=   0.3s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=700; total time=   0.5s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=900; total time=   0.8s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=900; total time=   0.8s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=900; total time=   0.7s
[CV] END criterion=squared_error, max_depth=3, max_features=log2, n_estimators=900; total time=   0.8s

{'criterion': 'squared_error',
 'max_depth': 2,
 'max_features': 'auto',
 'n_estimators': 200}


rf = GSCV.best_estimator_


rf.fit(Xtrain, Ytrain)

RandomForestRegressor(max_depth=2, n_estimators=200, random_state=30)


rf.predict(Xtest)

array([0.49937237, 0.51721498, 0.51721498, 0.50379367, 0.51299246,
       0.52611057, 0.51430736, 0.49776982, 0.48724366, 0.49407495,
       0.49814568, 0.51351549, 0.49863203, 0.51181923, 0.51315862,
       0.49911694, 0.5019478 , 0.5090705 , 0.49911694, 0.51493776,
       0.51299246, 0.50855023, 0.51210569, 0.51721498, 0.49806138,
       0.51138032, 0.46948417, 0.51250217, 0.45023305, 0.45310457,
       0.49776982, 0.51105479, 0.51243288, 0.49785246, 0.51721498,
       0.50760882, 0.44862968, 0.50399254, 0.45310457, 0.51240961,
       0.51721498, 0.51581007, 0.51721498, 0.48825309, 0.49687873,
       0.49482606, 0.51315862, 0.50400678, 0.5190723 , 0.49830911,
       0.49994347, 0.51315862, 0.49994347, 0.51250217, 0.4692275 ,
       0.44939456, 0.49512647, 0.51243698, 0.51318389, 0.49893996,
       0.49785246, 0.51430736, 0.45235154, 0.49512647, 0.5146737 ,
       0.44472791, 0.48596412, 0.51316999, 0.51430736, 0.46881999,
       0.50133668, 0.51206357, 0.48888741, 0.51181923, 0.5190723 ,
       0.51320038, 0.51181923, 0.46948417, 0.50186822, 0.51243698,
       0.51243288, 0.51721498, 0.51206357, 0.50855023, 0.50399254,
       0.51430736, 0.49806138, 0.44566076, 0.51299246, 0.46182183,
       0.51181923, 0.51243698, 0.47859821, 0.46948417, 0.44939456,
       0.51430736, 0.44472791, 0.44777585, 0.49333712, 0.51243288,
       0.51493776, 0.51240961, 0.51689932, 0.51206357, 0.48332848,
       0.51493776, 0.5019478 , 0.51299246, 0.44939456, 0.51181923,
       0.49994347, 0.50756467, 0.51293218, 0.51351549, 0.49407495,
       0.49994347, 0.51210569, 0.44939456, 0.51275006, 0.44538118,
       0.51318389, 0.51243288, 0.51609429, 0.51243288, 0.49823889,
       0.51430736, 0.51250217, 0.44426185, 0.50328828, 0.50305909,
       0.51243698, 0.51105479, 0.51181923, 0.44825103, 0.51947087,
       0.51240961, 0.49753716, 0.51240961, 0.51240961, 0.51721498,
       0.51181923, 0.49370656, 0.51243698, 0.51721498, 0.51138032,
       0.49020005, 0.51947087, 0.44538118, 0.51243698, 0.51721498,
       0.49826836, 0.51315862, 0.48724366, 0.51240961, 0.44472791,
       0.51689932, 0.51206357, 0.51721498, 0.46182183, 0.51181923,
       0.51721498, 0.46948417, 0.50471445, 0.51706577, 0.46948417,
       0.51243698, 0.51430736, 0.51430736, 0.51320038, 0.50400678,
       0.48332848, 0.44426185, 0.51706577, 0.51689932, 0.44328962,
       0.47859821, 0.49863203, 0.49826836, 0.50855023, 0.51243698,
       0.49863203, 0.51181923, 0.51721498, 0.4896781 , 0.51721498,
       0.50443437, 0.49823889, 0.50760882, 0.46948417, 0.49806138,
       0.51181923, 0.51293218, 0.44889125, 0.51609429, 0.51659114,
       0.51315862, 0.51320038, 0.51206357, 0.51721498, 0.51581007,
       0.49863203, 0.51581007, 0.5190723 , 0.51250217, 0.51250217,
       0.48666789, 0.49806138, 0.4896781 , 0.5053085 , 0.47859821,
       0.51430736, 0.50855023, 0.49994347, 0.51240961, 0.49333712,
       0.51320067, 0.4896781 , 0.45040466, 0.51299246, 0.49994347,
       0.48965645, 0.51583716, 0.49595718, 0.49773783, 0.49806138,
       0.51293218, 0.5019478 , 0.51140265, 0.5019478 , 0.49776982,
       0.49338115, 0.51947087, 0.49407495, 0.51299246, 0.51181923,
       0.51430736, 0.51721498, 0.49806138, 0.51318389, 0.51299246,
       0.44580474, 0.49512647, 0.51609429, 0.51493776, 0.46948417,
       0.5044319 , 0.48724366, 0.4896781 , 0.48429615, 0.51210569,
       0.51545894, 0.5190723 , 0.51243288, 0.49826836, 0.51721498,
       0.51430736, 0.49863203, 0.51493776, 0.45310457, 0.51243698,
       0.49806138, 0.51659114, 0.52503191, 0.5044319 , 0.48017606,
       0.4692275 , 0.51430736, 0.51299246, 0.50855023, 0.49863203,
       0.51430736, 0.49298072, 0.45040466, 0.51243288, 0.49753716,
       0.51181923, 0.49863203, 0.57733241, 0.49806138, 0.51299246,
       0.49613043])


Ytest

431.0    0.140630
45.0     0.158646
803.0    0.583773
674.0    0.328720
38.0     0.606789
           ...   
608.0    0.543305
487.0    0.499028
62.0     0.655680
399.0    0.649320
652.0    0.539078
Name: Прочность при растяжении, МПа, Length: 281, dtype: float64


np.mean(np.abs(Ytest-rf.predict(Xtest)))

0.15034493944156616


np.mean(np.abs(Ytest-np.mean(Ytest)))

0.1488040656405075

df


Y = norm_df[['Соотношение матрица-наполнитель']]
X = norm_df.drop(['Соотношение матрица-наполнитель'], axis=1)


Xtrain, Xholdout, Ytrain, Yholdout = train_test_split(X, Y, test_size=0.3, random_state=17)
print(Xtrain.shape, Xholdout.shape)

(655, 12) (281, 12)


normalizer = tf.keras.layers.Normalization(axis=-1)
normalizer.adapt(np.array(X))
print(normalizer.mean.numpy())

[[0.5026951  0.4467639  0.5046643  0.49121633 0.5160587  0.37373248
  0.48864686 0.49570566 0.5211413  0.5117521  0.5022317  0.5137764 ]]


def plot_loss(history):
    plt.plot(history.history['loss'], label='loss')
    plt.plot(history.history['val_loss'], label='val_loss')
    plt.ylim([0, 10])
    plt.xlabel('Эпоха')
    plt.ylabel('MAE')
    plt.legend()
    plt.grid(True)


def build_and_compile_model(norm):
    model = keras.Sequential([
      norm,
      layers.Dense(40, activation='relu'),
      layers.Dense(40, activation='relu'),
      layers.Dense(2)
    ])

    model.compile(loss='mean_absolute_error',
                optimizer=tf.keras.optimizers.Adam(0.001))
    return model


dnn_model = build_and_compile_model(normalizer)
dnn_model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 normalization (Normalizatio  (None, 12)               25        
 n)                                                              
                                                                 
 dense (Dense)               (None, 40)                520       
                                                                 
 dense_1 (Dense)             (None, 40)                1640      
                                                                 
 dense_2 (Dense)             (None, 2)                 82        
                                                                 
=================================================================
Total params: 2,267
Trainable params: 2,242
Non-trainable params: 25
_________________________________________________________________


%%time
history = dnn_model.fit(
    Xtrain.values,
    Ytrain.values,
    epochs=100,
    verbose=1,
    validation_split = 0.2)

Epoch 1/100
17/17 [==============================] - 0s 7ms/step - loss: 0.4584 - val_loss: 0.3421
Epoch 2/100
17/17 [==============================] - 0s 2ms/step - loss: 0.2713 - val_loss: 0.2670
Epoch 3/100
17/17 [==============================] - 0s 2ms/step - loss: 0.2177 - val_loss: 0.2258
Epoch 4/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1915 - val_loss: 0.2086
Epoch 5/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1784 - val_loss: 0.2023
Epoch 6/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1701 - val_loss: 0.1967
Epoch 7/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1630 - val_loss: 0.1953
Epoch 8/100
17/17 [==============================] - 0s 1ms/step - loss: 0.1574 - val_loss: 0.1929
Epoch 9/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1526 - val_loss: 0.1906
Epoch 10/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1490 - val_loss: 0.1876
Epoch 11/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1451 - val_loss: 0.1862
Epoch 12/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1422 - val_loss: 0.1864
Epoch 13/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1383 - val_loss: 0.1843
Epoch 14/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1364 - val_loss: 0.1829
Epoch 15/100
17/17 [==============================] - 0s 1ms/step - loss: 0.1320 - val_loss: 0.1826
Epoch 16/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1309 - val_loss: 0.1812
Epoch 17/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1281 - val_loss: 0.1810
Epoch 18/100
17/17 [==============================] - 0s 1ms/step - loss: 0.1278 - val_loss: 0.1800
Epoch 19/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1268 - val_loss: 0.1790
Epoch 20/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1247 - val_loss: 0.1792
Epoch 21/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1234 - val_loss: 0.1756
Epoch 22/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1205 - val_loss: 0.1762
Epoch 23/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1170 - val_loss: 0.1753
Epoch 24/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1152 - val_loss: 0.1740
Epoch 25/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1143 - val_loss: 0.1741
Epoch 26/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1121 - val_loss: 0.1732
Epoch 27/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1113 - val_loss: 0.1729
Epoch 28/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1087 - val_loss: 0.1745
Epoch 29/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1095 - val_loss: 0.1714
Epoch 30/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1067 - val_loss: 0.1717
Epoch 31/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1061 - val_loss: 0.1707
Epoch 32/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1030 - val_loss: 0.1719
Epoch 33/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1024 - val_loss: 0.1688
Epoch 34/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1014 - val_loss: 0.1696
Epoch 35/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1041 - val_loss: 0.1712
Epoch 36/100
17/17 [==============================] - 0s 2ms/step - loss: 0.1020 - val_loss: 0.1704
Epoch 37/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0985 - val_loss: 0.1688
Epoch 38/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0984 - val_loss: 0.1681
Epoch 39/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0960 - val_loss: 0.1709
Epoch 40/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0959 - val_loss: 0.1662
Epoch 41/100
17/17 [==============================] - 0s 5ms/step - loss: 0.0947 - val_loss: 0.1676
Epoch 42/100
17/17 [==============================] - 0s 6ms/step - loss: 0.0934 - val_loss: 0.1690
Epoch 43/100
17/17 [==============================] - 0s 5ms/step - loss: 0.0912 - val_loss: 0.1687
Epoch 44/100
17/17 [==============================] - 0s 5ms/step - loss: 0.0912 - val_loss: 0.1706
Epoch 45/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0933 - val_loss: 0.1720
Epoch 46/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0895 - val_loss: 0.1658
Epoch 47/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0866 - val_loss: 0.1678
Epoch 48/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0882 - val_loss: 0.1629
Epoch 49/100
17/17 [==============================] - 0s 5ms/step - loss: 0.0863 - val_loss: 0.1676
Epoch 50/100
17/17 [==============================] - 0s 4ms/step - loss: 0.0854 - val_loss: 0.1643
Epoch 51/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0827 - val_loss: 0.1685
Epoch 52/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0815 - val_loss: 0.1668
Epoch 53/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0814 - val_loss: 0.1681
Epoch 54/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0819 - val_loss: 0.1652
Epoch 55/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0787 - val_loss: 0.1655
Epoch 56/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0794 - val_loss: 0.1659
Epoch 57/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0800 - val_loss: 0.1658
Epoch 58/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0781 - val_loss: 0.1630
Epoch 59/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0771 - val_loss: 0.1663
Epoch 60/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0743 - val_loss: 0.1639
Epoch 61/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0750 - val_loss: 0.1662
Epoch 62/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0752 - val_loss: 0.1666
Epoch 63/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0741 - val_loss: 0.1661
Epoch 64/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0717 - val_loss: 0.1675
Epoch 65/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0717 - val_loss: 0.1687
Epoch 66/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0700 - val_loss: 0.1695
Epoch 67/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0689 - val_loss: 0.1687
Epoch 68/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0694 - val_loss: 0.1700
Epoch 69/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0684 - val_loss: 0.1691
Epoch 70/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0671 - val_loss: 0.1701
Epoch 71/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0672 - val_loss: 0.1721
Epoch 72/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0659 - val_loss: 0.1736
Epoch 73/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0670 - val_loss: 0.1689
Epoch 74/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0659 - val_loss: 0.1710
Epoch 75/100
17/17 [==============================] - 0s 3ms/step - loss: 0.0646 - val_loss: 0.1708
Epoch 76/100
17/17 [==============================] - 0s 3ms/step - loss: 0.0638 - val_loss: 0.1723
Epoch 77/100
17/17 [==============================] - 0s 4ms/step - loss: 0.0652 - val_loss: 0.1706
Epoch 78/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0612 - val_loss: 0.1713
Epoch 79/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0591 - val_loss: 0.1727
Epoch 80/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0614 - val_loss: 0.1719
Epoch 81/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0595 - val_loss: 0.1676
Epoch 82/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0608 - val_loss: 0.1732
Epoch 83/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0580 - val_loss: 0.1726
Epoch 84/100
17/17 [==============================] - 0s 3ms/step - loss: 0.0578 - val_loss: 0.1733
Epoch 85/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0564 - val_loss: 0.1731
Epoch 86/100
17/17 [==============================] - 0s 4ms/step - loss: 0.0568 - val_loss: 0.1758
Epoch 87/100
17/17 [==============================] - 0s 3ms/step - loss: 0.0566 - val_loss: 0.1735
Epoch 88/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0559 - val_loss: 0.1727
Epoch 89/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0540 - val_loss: 0.1752
Epoch 90/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0527 - val_loss: 0.1731
Epoch 91/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0532 - val_loss: 0.1764
Epoch 92/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0520 - val_loss: 0.1751
Epoch 93/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0536 - val_loss: 0.1723
Epoch 94/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0520 - val_loss: 0.1753
Epoch 95/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0529 - val_loss: 0.1760
Epoch 96/100
17/17 [==============================] - 0s 1ms/step - loss: 0.0526 - val_loss: 0.1770
Epoch 97/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0532 - val_loss: 0.1738
Epoch 98/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0538 - val_loss: 0.1760
Epoch 99/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0498 - val_loss: 0.1747
Epoch 100/100
17/17 [==============================] - 0s 2ms/step - loss: 0.0494 - val_loss: 0.1800
Wall time: 3.75 s


plot_loss(history)
dnn_model.evaluate(Xtest, Ytest, verbose=0)

0.22227410972118378


Y = norm_df[['Соотношение матрица-наполнитель']]
X = norm_df.drop(['Соотношение матрица-наполнитель'], axis=1)


Xtrain, Xholdout, Ytrain, Yholdout = train_test_split(X, Y, test_size=0.3, random_state=17)
print(Xtrain.shape, Xholdout.shape)

(655, 12) (281, 12)


normalizer2 = tf.keras.layers.Normalization(axis=-1)
normalizer2.adapt(np.array(X))
print(normalizer2.mean.numpy())

[[0.5026951  0.4467639  0.5046643  0.49121633 0.5160587  0.37373248
  0.48864686 0.49570566 0.5211413  0.5117521  0.5022317  0.5137764 ]]


def build_and_compile_model(norm):
    model2 = keras.Sequential([
      norm,
      layers.Dense(300, activation='sigmoid'),
        layers.Dropout(0.6),
      layers.Dense(20, activation='sigmoid'),
      layers.Dense(10, activation='sigmoid'),
      layers.Dense(1)
    ])

    model2.compile(loss='mean_absolute_error',
                optimizer=tf.keras.optimizers.RMSprop(0.001))
    return model2

dnn_model2 = build_and_compile_model(normalizer2)
dnn_model2.summary()

Model: "sequential_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 normalization_1 (Normalizat  (None, 12)               25        
 ion)                                                            
                                                                 
 dense_3 (Dense)             (None, 300)               3900      
                                                                 
 dropout (Dropout)           (None, 300)               0         
                                                                 
 dense_4 (Dense)             (None, 20)                6020      
                                                                 
 dense_5 (Dense)             (None, 10)                210       
                                                                 
 dense_6 (Dense)             (None, 1)                 11        
                                                                 
=================================================================
Total params: 10,166
Trainable params: 10,141
Non-trainable params: 25
_________________________________________________________________


%%time
history2 = dnn_model2.fit(
    Xtrain,
    Ytrain,
    validation_split = 0.2,
    verbose=0, epochs=50)

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<timed exec> in <module>

~\anaconda3\envs\learning\lib\site-packages\keras\utils\traceback_utils.py in error_handler(*args, **kwargs)
     65     except Exception as e:  # pylint: disable=broad-except
     66       filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67       raise e.with_traceback(filtered_tb) from None
     68     finally:
     69       del filtered_tb

~\anaconda3\envs\learning\lib\site-packages\pandas\core\frame.py in __getitem__(self, key)
   3476 
   3477         # Do we have a slicer (on rows)?
-> 3478         indexer = convert_to_index_sliceable(self, key)
   3479         if indexer is not None:
   3480             if isinstance(indexer, np.ndarray):

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexing.py in convert_to_index_sliceable(obj, key)
   2314     idx = obj.index
   2315     if isinstance(key, slice):
-> 2316         return idx._convert_slice_indexer(key, kind="getitem")
   2317 
   2318     elif isinstance(key, str):

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\numeric.py in _convert_slice_indexer(self, key, kind)
    275             # We always treat __getitem__ slicing as label-based
    276             # translate to locations
--> 277             return self.slice_indexer(key.start, key.stop, key.step)
    278 
    279         return super()._convert_slice_indexer(key, kind=kind)

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\base.py in slice_indexer(self, start, end, step, kind)
   6272         self._deprecated_arg(kind, "kind", "slice_indexer")
   6273 
-> 6274         start_slice, end_slice = self.slice_locs(start, end, step=step)
   6275 
   6276         # return a slice

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\base.py in slice_locs(self, start, end, step, kind)
   6482         start_slice = None
   6483         if start is not None:
-> 6484             start_slice = self.get_slice_bound(start, "left")
   6485         if start_slice is None:
   6486             start_slice = 0

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\base.py in get_slice_bound(self, label, side, kind)
   6401             except ValueError:
   6402                 # raise the original KeyError
-> 6403                 raise err
   6404 
   6405         if isinstance(slc, np.ndarray):

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\base.py in get_slice_bound(self, label, side, kind)
   6395         # we need to look up the label
   6396         try:
-> 6397             slc = self.get_loc(label)
   6398         except KeyError as err:
   6399             try:

~\anaconda3\envs\learning\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   3621                 return self._engine.get_loc(casted_key)
   3622             except KeyError as err:
-> 3623                 raise KeyError(key) from err
   3624             except TypeError:
   3625                 # If we have a listlike key, _check_indexing_error will raise

KeyError: 0


plot_loss(history)
dnn_model2.evaluate(Xtest, Ytest, verbose=0)

0.9249774813652039

	Соотношение матрица-наполнитель	Плотность, кг/м3	модуль упругости, ГПа	Количество отвердителя, м.%	Содержание эпоксидных групп,%_2	Температура вспышки, С_2	Поверхностная плотность, г/м2	Модуль упругости при растяжении, ГПа	Прочность при растяжении, МПа	Потребление смолы, г/м2
0.0	1.857143	2030.000000	738.736842	30.000000	22.267857	100.000000	210.000000	70.000000	3000.000000	220.000000
1.0	1.857143	2030.000000	738.736842	50.000000	23.750000	284.615385	210.000000	70.000000	3000.000000	220.000000
2.0	1.857143	2030.000000	738.736842	49.900000	33.000000	284.615385	210.000000	70.000000	3000.000000	220.000000
3.0	1.857143	2030.000000	738.736842	129.000000	21.250000	300.000000	210.000000	70.000000	3000.000000	220.000000
4.0	2.771331	2030.000000	753.000000	111.860000	22.267857	284.615385	210.000000	70.000000	3000.000000	220.000000
...	...	...	...	...	...	...	...	...	...	...
1018.0	2.271346	1952.087902	912.855545	86.992183	20.123249	324.774576	209.198700	73.090961	2387.292495	125.007669
1019.0	3.444022	2050.089171	444.732634	145.981978	19.599769	254.215401	350.660830	72.920827	2360.392784	117.730099
1020.0	3.280604	1972.372865	416.836524	110.533477	23.957502	248.423047	740.142791	74.734344	2662.906040	236.606764
1021.0	3.705351	2066.799773	741.475517	141.397963	19.246945	275.779840	641.468152	74.042708	2071.715856	197.126067
1022.0	3.808020	1890.413468	417.316232	129.183416	27.474763	300.952708	758.747882	74.309704	2856.328932	194.754342

	Угол нашивки, град	Шаг нашивки	Плотность нашивки
0.0	0.0	4.000000	57.000000
1.0	0.0	4.000000	60.000000
2.0	0.0	4.000000	70.000000
3.0	0.0	5.000000	47.000000
4.0	0.0	5.000000	57.000000
...	...	...	...
1035.0	90.0	8.088111	47.759177
1036.0	90.0	7.619138	66.931932
1037.0	90.0	9.800926	72.858286
1038.0	90.0	10.079859	65.519479
1039.0	90.0	9.021043	66.920143

	Соотношение матрица-наполнитель	Плотность, кг/м3	модуль упругости, ГПа	Количество отвердителя, м.%	Содержание эпоксидных групп,%_2	Температура вспышки, С_2	Поверхностная плотность, г/м2	Модуль упругости при растяжении, ГПа	Прочность при растяжении, МПа	Потребление смолы, г/м2	Угол нашивки, град	Шаг нашивки	Плотность нашивки
0.0	1.857143	2030.000000	738.736842	30.000000	22.267857	100.000000	210.000000	70.000000	3000.000000	220.000000	0.0	4.000000	57.000000
1.0	1.857143	2030.000000	738.736842	50.000000	23.750000	284.615385	210.000000	70.000000	3000.000000	220.000000	0.0	4.000000	60.000000
2.0	1.857143	2030.000000	738.736842	49.900000	33.000000	284.615385	210.000000	70.000000	3000.000000	220.000000	0.0	4.000000	70.000000
3.0	1.857143	2030.000000	738.736842	129.000000	21.250000	300.000000	210.000000	70.000000	3000.000000	220.000000	0.0	5.000000	47.000000
4.0	2.771331	2030.000000	753.000000	111.860000	22.267857	284.615385	210.000000	70.000000	3000.000000	220.000000	0.0	5.000000	57.000000
...	...	...	...	...	...	...	...	...	...	...	...	...	...
1018.0	2.271346	1952.087902	912.855545	86.992183	20.123249	324.774576	209.198700	73.090961	2387.292495	125.007669	90.0	9.076380	47.019770
1019.0	3.444022	2050.089171	444.732634	145.981978	19.599769	254.215401	350.660830	72.920827	2360.392784	117.730099	90.0	10.565614	53.750790
1020.0	3.280604	1972.372865	416.836524	110.533477	23.957502	248.423047	740.142791	74.734344	2662.906040	236.606764	90.0	4.161154	67.629684
1021.0	3.705351	2066.799773	741.475517	141.397963	19.246945	275.779840	641.468152	74.042708	2071.715856	197.126067	90.0	6.313201	58.261074
1022.0	3.808020	1890.413468	417.316232	129.183416	27.474763	300.952708	758.747882	74.309704	2856.328932	194.754342	90.0	6.078902	77.434468

	Соотношение матрица-наполнитель	Плотность, кг/м3	модуль упругости, ГПа	Количество отвердителя, м.%	Содержание эпоксидных групп,%_2	Температура вспышки, С_2	Поверхностная плотность, г/м2	Модуль упругости при растяжении, ГПа	Прочность при растяжении, МПа	Потребление смолы, г/м2	Угол нашивки, град	Шаг нашивки	Плотность нашивки
count	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000	1023.000000
mean	2.930366	1975.734888	739.923233	110.570769	22.244390	285.882151	482.731833	73.328571	2466.922843	218.423144	44.252199	6.899222	57.153929
std	0.913222	73.729231	330.231581	28.295911	2.406301	40.943260	281.314690	3.118983	485.628006	59.735931	45.015793	2.563467	12.350969
min	0.389403	1731.764635	2.436909	17.740275	14.254985	100.000000	0.603740	64.054061	1036.856605	33.803026	0.000000	0.000000	0.000000
25%	2.317887	1924.155467	500.047452	92.443497	20.608034	259.066528	266.816645	71.245018	2135.850448	179.627520	0.000000	5.080033	49.799212
50%	2.906878	1977.621657	739.664328	110.564840	22.230744	285.896812	451.864365	73.268805	2459.524526	219.198882	0.000000	6.916144	57.341920
75%	3.552660	2021.374375	961.812526	129.730366	23.961934	313.002106	693.225017	75.356612	2767.193119	257.481724	90.000000	8.586293	64.944961
max	5.591742	2207.773481	1911.536477	198.953207	33.000000	413.273418	1399.542362	82.682051	3848.436732	414.590628	90.000000	14.440522	103.988901

	Соотношение матрица-наполнитель	Плотность, кг/м3	модуль упругости, ГПа	Количество отвердителя, м.%	Содержание эпоксидных групп,%_2	Температура вспышки, С_2	Поверхностная плотность, г/м2	Модуль упругости при растяжении, ГПа	Прочность при растяжении, МПа	Потребление смолы, г/м2	Угол нашивки, град	Шаг нашивки	Плотность нашивки
count	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000	936.000000
mean	0.498933	0.502695	0.446764	0.504664	0.491216	0.516059	0.373733	0.488647	0.495706	0.521141	0.511752	0.502232	0.513776
std	0.187489	0.187779	0.199583	0.188865	0.180620	0.190624	0.217078	0.191466	0.188915	0.195781	0.500129	0.183258	0.191342
min	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	0.372274	0.368517	0.301243	0.376190	0.367716	0.386128	0.205619	0.359024	0.365149	0.392067	0.000000	0.372211	0.390482
50%	0.494538	0.511229	0.447061	0.506040	0.489382	0.515980	0.354161	0.485754	0.491825	0.523766	1.000000	0.504258	0.516029
75%	0.629204	0.624999	0.580446	0.637978	0.623410	0.646450	0.538683	0.615077	0.612874	0.652447	1.000000	0.624604	0.638842
max	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000

ВЫПУСКНАЯ КВАЛИФИКАЦИОННАЯ РАБОТА по курсу «Data Science»¶

Слушатель Саморукова Юлия Дмитриевна¶

1. Изучить теоретические основы и методы решения поставленной задачи¶

Предоставленная информация:¶

2. Провести разведочный анализ предложенных данных¶

Импорт библиотек¶

Производим загрузку данных с локального диска.¶

Объединяю таблицы с помощью Inner¶

Часть строк из таблицы X_nup была исключена, в датасете осталось 1023 строки с 13 признаками¶

Средние и медианные значения¶

Проверка пропусков и дубликатов¶

Тип данных с плав.запятой, строковых значений не имеет. Пропущенные значения отсуствуют.¶

Гистограммы распределения каждой из переменной¶

Диаграммы размаха (ящик с усами)¶

Попарные графики рассеяния точек¶

На графике рассеяния видны небольшие выбросы.¶

Матрица корреляции данных¶

3) Провести предобработку данных (удаление шумов, нормализация и т.д.)¶

Признаки с повторениями отсутствуют. Убираем выбивающиеся из распределения значения.¶

Нормализация¶

4) Обучение моделей для прогноза модуля упругости при растяжении и прочности при растяжении¶

Модель линейной регрессии для параметра прочность при растяжении¶

Тестовое значение R^2 принимает отрицательное число.¶

Вывод: отсутствие линейной взаимосвязи в модели¶

Метод K-ближайших соседей для параметра прочность при растяжении¶

Random forest для параметра прочность при растяжении¶

Результат на глубину в 10 слоев и 1500 итераций - значение коэффициента детерминации все также отрицательное.¶

Для параметра упругость¶

Модель линейной регрессии для параметра упругость при растяжении¶

Метод K-ближайших соседей для упругости при растяжении¶

Random forest для параметра упругость при растяжении¶

Поиск Гиперпараметров¶

Импортируем функции для поиска по сетке параметров с перекрестной проверкой¶

Сравнение ошибки базовой модели¶

Многослойный персептрон¶

5.Нейронная сеть, которая будет рекомендовать соотношение матрица-наполнитель¶

Нейронная сеть¶

	модуль упругости, ГПа
775.0	0.301010
370.0	0.483233
50.0	0.243451
311.0	0.326373
868.0	0.512460
...	...
846.0	0.677026
78.0	0.386467
994.0	0.462350
263.0	0.493137
40.0	0.275313